Using Machine Learning to Optimize Graph Execution on NUMA Machines
TimeThursday, July 14th1:30pm - 1:53pm PDT
Location3004, Level 3
Event Type
Research Manuscript
Embedded Software
Embedded Systems
DescriptionThis paper proposes PredG, a Machine Learning framework to improve the graph processing performance by finding the ideal thread and data mapping on NUMA systems. PredG is agnostic to the graphs: it uses the graphs' features to train an ANN to perform predictions as new graphs arrive -- without any application execution after being trained. When evaluating PredG over representative graphs and algorithms on two NUMA systems, its solutions are up to 23% and 19% faster than the Linux OS Default and the Best Static - at most 6% far from the Oracle -, and it presents lower energy consumption.